Multiple Myeloma (MM) is an incurable disease in which malignant plasma cells proliferate excessively in the bone marrow (BM). Most patients initially respond well to therapy, which can extend survival for several years. However, these patients inevitably become refractory to therapy, resulting in relapse and death. Thus, persistence of MM cells following initial therapy leads to evolution of resistance and is the major cause of treatment failure. Development of MM is the result of complex tumor-host interactions in the bone marrow. These interactions also protect the MM cells from immune response and chemotherapy?termed cell-adhesion mediated drug resistance (CAM-DR). Our group has recently shown that relapsed patients (n=7) show significantly higher levels of CAM-DR markers (levels of a4 integrin in CD138+ MM cells from bone marrow aspirates) than observed in newly diagnosed patients (n=7). This strongly suggests that MM cells adapt to therapy at least in part by upregulating adhesion-mediated interactions. MM response to therapy can be evaluated using in vitro co-culture models. However, these data are limited to fixed time-points and drug concentrations, so that temporal and spatial heterogeneity that typically occur in the phenotype of MM cells and the BM environment are not examined. Thus, while traditional in vitro techniques are necessary for demonstrating drug efficacy and mechanisms, they are insufficient to fully characterize the range of cell response in vivo due to variation in CAM-DR, disordered vascularity and blood flow. Thus, factors likely to affect the outcome of therapy in individual patients (such as gradients of drugs and oxygen, or variability in regional density of bone marrow mesenchymal or adhesion molecules) have not, until now, been included in pre-clinical evaluation of MM therapies. In this study we propose an integrated computational/experimental approach that better recreates the temporal and spatial BM heterogeneity, utilizing a novel micro-fluidic device that can generate gradients of chemotherapeutic drugs and CAM-DR dynamics. These data will be used to parameterize computational models of the 3D BM structure, capable of reproducing the evolution of CAM-DR and therapy response. Preliminary validation of this proof of concept will consist in treatment simulations of personalized patient BM computational models, and comparing the model predictions to obtained BM biopsies. The end point of this exploratory grant is the proof-of-concept of this approach, namely a set of preliminary results from a small group of patients (n=9, split in three different stages of the disease), a customizable computational model of patient myelomatous bone marrow, and an optimized protocol for use of patient samples, immunohistochemistry and microfluidics for generating these models, which we plan to use in the future to support a clinical trial as an ancillary biomarker.